Re: URN namespace

Keith Moore (moore@cs.utk.edu)
Wed, 16 Aug 1995 17:11:34 -0400


Message-Id: <199508162111.RAA25062@wilma.cs.utk.edu>
From: Keith Moore <moore@cs.utk.edu>
To: mshapiro@ncsa.uiuc.edu (Michael Shapiro)
Cc: moore@cs.utk.edu (Keith Moore), uri@bunyip.com
Subject: Re: URN namespace 
In-Reply-To: Your message of "Wed, 16 Aug 1995 07:36:53 CDT."
             <9508161236.AA21785@void.ncsa.uiuc.edu> 
Date: Wed, 16 Aug 1995 17:11:34 -0400

> I was trying to ask these 2 questions:
> 
> 1. To get long term persistence, do we have to have a new name space?
> 2. Will a new top level DNS domain be a sufficient new name space, or
>    do we have to do something which is not DNS based?
> 
> Your answer to (1) is that we do. I think so, too. I thought Karen
> Sollins was saying this as well.  I'd like Karen and others besides you
> and I give an answer.
> 
> Answers to (2) are wanting.  DNS is very good for resolution.
> (I have some timings on "path" lookups that will pleasantly suprise you.)

DNS works well when it succeeds and when all the servers are up and
running; it doesn't work so well when one or more servers is down and
unreachable.  This gets worse as the number of different servers you
have to talk to in order to complete a query increases.  The maximum
amount of time that you have to wait before being sure that a query
won't be answered can be quite large.

A lot of this may not be due to DNS itself, but to the quirks of some
implementations (that mysteriously fail to answer queries sometimes)
and the way that DNS servers are managed.

> Is DNS good enough wrt name assignment?  It might be that the way names
> get created in DNS might not be good enough to help insure persistence.

To get persistent DNS names for use in URNs, we need them to be opaque
(not human-meaningful) and we need to make sure that the resolution
servers for those names can move around without changing the name.  I
don't see that DNS constrains either of those, but it might be
necessary to have various "neutral" parties run the DNS servers for
the top levels of that chunk of DNS space.

Another thing we may want to avoid is excessive fanout like that which
has plagued the .COM domain.  To me this means having one or two
layers of sub-domains below the .DNS top-level domain, but above that
delegated to any publisher.  It might also mean limiting the fanout at
any level to some fixed number.

Another characteristic we want is for nobody to be able to exercise
control over DNS space, and thus prevent someone from publishing.
This means at the very least that there should be multiple places from
which to get a chunk of DNS space for use in URNs.  (It's less clear
what to do about this in the case where the entity that maintains your
parent domain wants to jerk you around *after* you are already using
your chunk of DNS space.)

I suspect, for various reasons having to do with costs and
responsibility, that you'll have to pay some sort of signup and/or
maintenance fees to your parent domain for your chunk of .URN space,
but you'll get better service -- including the right to expect a
certain level of availability of those DNS servers, the right to sue
if they stop referring your queries to you, etc.

> I have heard statements that it is hard to get a new top level DNS
> domain.  I'm not sure this is true, but even if it were, should that be
> a guiding principle - ease of creating and extending the name space?

I suspect that if we get WG and IESG approval for the idea, that it
won't be a problem to actually create the new top-level domain.  In
other words: it may be difficult to get such approval.  The IESG can
be expected to scrutinize such a proposal very carefully, but I think
it's safe to assume that they want us to solve the URN problem
somehow.  It's up to us to make the best proposal we can and let
interested parties evaluate it and compare it with other ways of doing
things.  If the DNS solution looks like the best way to go, they'll
use it.

> Does ease (or difficulty) of creating and extending the name space
> affect persistence?  If it is easy to create a naming authority, will
> this work against persistence?

I think there's something to be said for requiring some investment in
effort to get a naming authority created in URN space.  Providing
persistent name service will require a commitment of resources that
not everyone is capable of making, and there's little benefit in
handing out chunks of URN space to those not willing to make such a
committment.  That said, if you have to go through a formal
registration procedure and fork over some money to cover maintenance
costs, the barrier is probably high enough.

Keith